Specifying magnetic bearings

Turbomachinery International, Jan/Feb 2007 by Richard, Philipp, Larsonneur, Ren�

APPLICATION AND IMPACT OF NEW INDUSTRY STANDARD ISO 14839

In the last ten years, active magnetic bearings have come out of the laboratory and become established alternatives to high-performance bearings. As their use becomes more widespread, the turbomachinery industry has been asking for guidelines and standards for design, acceptance and operation of rotating machines equipped with magnetic bearings. Although standards, such as API 617, cover a broad range of topics related to bearings in turbomachinery applications, there is a need for a more specific code focusing on the unique properties and capabilities of magnetic bearings.

The ISO 14839 standard (Table 2), covering vibration, stability and robustness for magnetic bearings used in turbomachines, could satisfy this need. And it could help evolve an industry consensus on the requirements for the performance of these systems, deepen user understanding of magnetic bearing technology, and improve user-supplier relationships.

Insufficient codes

Several standards are in use today that deal with rotordynamics in turbomachines (Table 1). The best known is the API 617, which covers applications such as centrifugal, axial and integrally geared compressors, as well as expander-compressor systems. API 617 provides guidelines for the design, manufacturing, modelling and testing of these systems.

All these standards emerged from the needs of classical turbomachines equipped with oil bearings rather than magnetic bearings. Consequently, they do not account for the unique properties of magnetic bearing-equipped machines, which are fundamentally different from those of machines with oil bearings. For instance, components of magnetic bearing systems are not subject to wear and fatigue when exposed to high vibration levels from unbalance.

In fact, magnetic bearing-equipped machines are suitable for unlimited, reliable and safe operation even in the presence of "large" residual unbalance levels. By allowing the rotor shaft to rotate about its principal axis, they have the ability to nearly eliminate bearing reaction forces induced by unbalance. For this reason, the laborious and expensive process of establishing and verifying residual unbalance levels, as extensively described in API 617 and ISO 1940, is unnecessary and "obsolete" for magnetic bearings.

Annex 4F of API 617, which is dedicated to magnetic bearing-equipped machines, does address this issue. The annex specifies a maximum allowable rotor movement relative to the center of the auxiliary bearing to assess unbalance vibration, and clarifies that "... this criterion supersedes all other vibration acceptance criteria as described for oil bearing machines ..." While this is correct, there are other aspects of magnetic bearing systems, such as stability and robustness, that API 617 does not consider.

Magnetic bearings are actively controlled systems that rely on the feedback of the rotor position or other measurable system states and, therefore, represent a closed-loop system architecture. The rotor, the bearings, sensors and power electronics constitute a dynamic system with properties that can be set by designing a controller with the required closed-loop stability and robustness.

To date, the term "robustness" of a closed-loop system has not received adequate attention and is sometimes even equated with the term "stability." Typically, a controller is designed as per a mathematical model that simplifies a "real-world" situation. But no real physical system truly behaves like the differential equations used in the math. Therefore, controllers must be "robust" so that their properties do not change significantly if applied to a system slightly different from the mathematical one.

A closed-loop system can be "very stable," but at the same time, "not robust," i.e., it can be sensitive to changes in its nominal parameters. For instance, a rotor shaft whose bending mode is actively damped by magnetic bearings can be "very stable," depending on the amount of damping introduced. However, the system can become unstable if the resonance frequency changes "slightly," which could occur due to thermal effects on rotor shafts.

While closed-loop stability can be assessed by criteria such as amplification factors, i.e., the sharpness of vibration peaks near resonance frequencies, there has, to date, not been an appropriate measure to assess system robustness. Part 3 of the new ISO 14839 standard, which is available as a draft, addresses this problem by introducing a "sensitivity function" to measure system robustness. This approach is based on the latest control theory concepts.

Parts 1 and 2 of ISO 14839 incorporate Annex 4F of API 617. Assessment criteria and vibration zone definitions are similar, thus easing transition from the API to the ISO framework. However, ISO 14839-3 is self-contained and exclusively addresses system robustness.

New criteria

In control theory, the "Nyquist" criterion has long been used to assess a system's stability margin. But this criterion is typically under a Single Input Single Output (SISO) framework. ISO 14839-3 generalizes this concept and extends it to a Multiple Input Multiple Output (MIMO) framework to measure robustness.